Warning message: optimization attribute on ‘Vec2d round(Vec2d)’ follows definition but the attribute doesn’t match [-Wattributes]"
Work around: Use option -Wno-attributes
Fixed 2021
Bug tracing: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89325
Intrinsic function _mm256_cmp_epi8_mask returns 16 bit mask, should be 32 bit.
A fix appears to be underway.
Work around in vectori256.h function _mm256_cmp_epi8_mask_fix
Bug doc: https://patchwork.ozlabs.org/patch/940760/
Gcc does not support the __vectorcall calling convention in Windows, needed in vectormath_lib.h
Work around in vectormath_lib.h: __ attribute __((sysv_abi))
__vectorcall could speed up all non-inline vector functions in 64-bit Windows.
Bug tracing: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89485
Gcc returns 256- and 512-bit vectors through a pointer in Win64, while all other compilers use a register.
This can cause errors when using function libraries or when different parts of a program are compiled with different compilers.
Bug tracing: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89597
This is not a bug, only suboptimal code.
Bug tracing: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87767
Fixed 2021 in Gcc version 11 and 12
This bug occurs when compiling for Windows.
Work around: Compile with option -fno-asynchronous-unwind-tables
Bug tracing: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65782
Fixed in Gcc 8 and 9, February 2020.